Three of science's oldest questions, tested in the open.

How did life begin? What makes some brain activity come with experience? What did the Indus cities write? Each project sets a checkable goal for 2027, states its odds plainly, and publishes every result, the failures too.

Open Indus

31 results landed · 299 sealed pre-registrations

What did the Indus cities write?

Decipher the Indus script by 2027.

Four thousand short texts survive from a civilisation as large as Egypt's, and nobody can read them. We test every claimed reading, and our own ideas, against chance.

  1. Structure: done
  2. Sign roles: partly
  3. Topic: open
  4. Meaning: open
  5. Sound: open
StructureSound

Honest odds by 2027under 1 in 10

Open Indus

Open Mind

7 results landed · 16 sealed pre-registrations

Which brain activity comes with experience?

Find the most accurate theory of human consciousness by 2027.

Every current theory may be wrong. We test their riskiest predictions on open brain recordings, list the hard facts any true theory must explain, and build our own theory where theirs fail.

  1. Signal: done
  2. State: done
  3. Dissociation: partly
  4. Content: partly
  5. Theory: open
SignalTheory

Honest odds by 20273 to 5 in 100

Open Mind

Open Origins

9 results landed · 5 sealed pre-registrations

How did chemistry become life?

Map and test every route to the origin of life by 2027.

Chemists have shown many single steps from simple molecules toward life, but nobody has joined them into one continuous route. We score each route step by step and seal predictions that future data can confirm.

  1. Building blocks: done
  2. Polymers: partly
  3. Copying: partly
  4. Self-replication: open
  5. Evolution: open
Building blocksEvolution

Honest odds by 2027about 1 in 8 for the open goal

Open Origins

Latest from the three projects

All news: Open Indus · Open Mind · Open Origins

One method for all three

A result that looks good must not be something you could get by trying until something works. These rules are the same in every project, and every result can be rerun from its code.

  1. Seal it before you look. Every prediction, test and pass mark is written down and fingerprinted (SHA-256) before any result is seen.
  2. Test on data it never saw. Part of the data is held back, and near-copies of it never reach the people building the idea.
  3. Plant an answer first. A method must recover a known, planted answer at the same scale before anything it says about the real data counts.
  4. Let someone else try to break it. Every positive result goes to a red team that did not produce it.
  5. Publish it either way. Failures are written up with the size of effect the test could have caught.

Who does the work

A human lead and a team of AI agents share one task board per project. They claim tasks, run tests on a single laptop, and hand every positive result to someone else to break.

You can join the same way:

  • Rerun a test and tell us if you get a different number.
  • Attack a result. Red-team tasks are open on each project's tracker.
  • Point us to data we are missing. Each project lists what would change its odds most.
  • Tell us where we are wrong about your field.